Use of antidepressant medications by older adults has increased significantly over the past 10 years. This increase, however, has not been distributed equally by race/ethnicity. To explore antidepressant use in older adults, we propose additional analyses of data from the Duke Established Populations for Epidemiological Studies of the Elderly (EPESE) sample (4,162 community dwelling elders initially interviewed in 1986 and followed over ten years via three additional in person and four telephone follow-up interviews). [unreadable] [unreadable] Detailed current medication use was obtained from sample members during each in-person interview. In controlled analyses, Whites were two to four times as likely to be taking antidepressant medications as African Americans at each survey and the differences by race increased over a ten year period of follow-up. This difference was especially noted between years six and ten of the follow-up study, with 80% of new users of antidepressant medications over this period being White. We propose a model to determine propensities to use antidepressants in Whites and to test the concordance/discordance of this propensity model in African Americans. This model includes eight domains: demographics; education/occupation/economic well being; geographic characteristics; health and mental health; personal attitudes; social and spiritual resources; health behaviors; and physician provider characteristics. The richness of this model derives from additional questionnaire data, HCFA Part A and Part B files, geocoding the residence of sample members, and detailed descriptions of the physicians listed as their primary providers. Hypotheses are proposed to construct propensity scores among Whites for the eight domains including: 1) there will be a decreased propensity to take antidepressants among participants whose primary source of care is a primary care physician who practices in a rural county; and 2) a perceived impairment in the social network will increase the propensity to use antidepressants. Our main hypothesis is that the propensity model derived for Whites will be discordant for African Americans and the discordance will be found in factors for Personal Mastery, spiritual resources, alcohol and medication use, and racial concordance. We will also test hypotheses specific to individual domains in the larger model.